Computers and Structures 80 (2002) 515531

www.elsevier.com/locate/compstruc

Post-limit stiness and ductility of end-plate

beam-to-column steel joints
~es da Silva
L. Simo

a,*

, Aldina Santiago b, Paulo Vila Real

Department of Civil Engineering, University of Coimbra, Polo II, Pinhal de Marrocos, 3030 Coimbra, Portugal
Department of Civil Engineering, University of Beira Interior, Edifcio II das Engenharias, Calcada do Lameiro, 6200 Covilh~
a, Portugal
c
Department of Civil Engineering, University of Aveiro, Campo de Santiago, 3800 Aveiro, Portugal
Received 26 July 2001; accepted 9 January 2002

Abstract
A procedure for the evaluation of ductility in steel joints is presented. Using the component method as background,
a non-linear analysis for a number of end-plate beam-to-column joints is performed that is capable of identifying the
yield sequence of the various components and the failure of the joint. Each component is characterised using a bilinear approximation for the forcedisplacement relation. Comparing these results with the corresponding experimental
results leads to a proposal of the post-limit stiness of the various components. A component ductility index is proposed for each component as a means of classication with respect to ductility, using the three ductility classes currently
proposed in the literature. A joint ductility index is also proposed, which can be used to verify available rotation against
the structure required rotation. 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Steel joints; Steel structures; Ductility; Component method; Post-limit stiness

1. Introduction
It is never enough repeating that the behaviour of
joints is complex, falling between the traditional assumption of pinned or fully rigid response. A considerable eort was undertaken over the past two decades
to give consistent predictions of the behaviour of steel
joints. However, until now, most research studies on the
behaviour of semi-rigid joints were focused on determining resistance and stiness characteristics [8,17,19]
leading, for example, to the code specications for the
evaluation of strength and stiness of steel joints that
were prepared for Eurocode 3 [2].
The evaluation of joint ductility constitutes an essential characteristic to ensure that sucient rotation or
deformation capacity is available to allow the chosen

Corresponding author. Tel.: +351-239-797216; fax: +351239-797217.

E-mail address: luis_silva@gipac.pt (L. Sim~
oes da Silva).

analysis type (elastic, plastic). Fig. 1 illustrates the momentrotation response of a very sti, overstrength joint
(rigid in practical terms) that, for a given applied moment M, exhibits a rotation h0j < hb , hb denoting the
corresponding beam rotation. In contrast, for the same
applied moment, a semi-rigid joint will reach a rotation
h00j > hb , thus requiring much higher ductility from the
joint. This ductility demand may easily reach 0.03 rad
for some joints where a plastic hinge is required, in
plastic design conditions.
The prediction of the deformation of beam-tocolumn or beam-to-beam steel joints requires the consideration of bending moments and axial and shear forces,
that are usually present in a steel frame. Concentrating solely on the rotational deformations arising from
bending moments of the connected beam, it is necessary
to dene the various contributions for the total rotation
of the joint. In analogy with member rotation, and given
that non-linearity in the momentrotation response of
steel connections starts at low values of rotation, ductility ratios are proposed in this paper, aimed at the

0045-7949/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 5 - 7 9 4 9 ( 0 2 ) 0 0 0 1 4 - 7

516

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

Nomenclature
A
As
Ac
At
Aw
Awc
Bt;Rd
E
Fy
Fb;fc;Rd
Fb;p;Rd
Fc;fb;Rd

Fc;wc;Rd
Ft;wb;Rd
Ft;wc;Rd
Fv;Rd
G
Ke
K pl
Kb;f
Kb;p
Kc;wc
Kt;b
Kt;wb
Kt;wc
Kw
Kwp
Lb

M
Mc;Rd
Mpl;Rd
Q
Vwc;Rd
a
b
beff
beff;c;wc

cross-section area
tensile stress area of the bolt
eective web area in compression zone
eective web area in tension zone
eective shear area
shear area of the column
tensile resistance of bolts
Youngs modulus
strength of component
resistance of column ange in bending
resistance of end-plate in bending
compression resistance of a beam ange and
the adjacent compression zone of a beam
web
resistance of an unstiened column web
subject to transverse compression
tensile resistance of the beam web
tensile resistance of the column web
shear resistance of bolts
shear modulus of steel
elastic stiness of the connection, obtained
from the following: ki E
post-limit stiness of the connection
stiness of component column ange in
bending
stiness of component end-plate in bending
stiness of component column web in compression
stiness of component bolt in tension
stiness of component beam web in tension
stiness of component column web in tension
stiness of component weld
stiness of component column web panel in
shear
bolt elongation length, taken as equal to the
grip length (total thickness of material and
washers), plus half the sum of the height of
the bolt head and the height of the nut
moment
moment resistance of the beam cross-section, reduced if necessary to allow for shear
exural resistance of end-plate or column
ange
shear force of the column web
shear resistance of column web panel
eective thickness of the weld
ange width
eective width
eective width of the column in compression
zone

eective width of the beam web in tension

zone
beff;t;wc eective width of the column web in tension
zone
db
clear depth of the beam
dc
clear depth of the column
fu
ultimate tensile strength of the weld
fub
ultimate tensile strength of the bolts
fy
yield stress
fy;p
yield stress of end-plate
fy;wb
yield stress of a beam web
fy;wc
yield stress of a column web
h
depth of the column
ht
distance from the tensile force to the centre
of compression
leff
smallest of the eective lengths (individually
or as part of a bolt group)
ki
stiness coecients for basic joint components
m
distance between the bolt centre-line and the
face of the weld connecting the beam web to
the end-plate; number of faying surfaces or
shear planes in a bolted joint, equal to 1.0
for bolts in single shear and 2.0 for bolts in
double shear
n
eective distance to the free edge
nb
number of bolt-rows
r
root radius of the web-ange junction
tf
ange thickness
tfb
thickness of a beam ange
tfc
thickness of a column ange
tp
thickness of end-plate
tw
web thickness
twb
thickness of a beam web
twc
thickness of a column web
z
lever arm between the compressive resistance, and tensile resistance
m
Poissons ratio
b
transformation parameter, see Part 18,
EC3
bw
correlation factor
h
rotation
h0j ; h00j ; hb design rotation to rigid joint; semi-rigid
joint and beam, respectively
x
reduction factor to allow for the possible
eects of shear in the column web panel
q
reduction factor for plate buckling
Dy
deformation that correspond to yield stress
Df
deformation that correspond to failure
Di
deformation for basic joint components
beff;t;wb

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

517

2. Joint models
2.1. Component method

Fig. 1. Comparison of momentrotation response between

beam and various joint types.

objective of ensuring ductility compliance, within the

framework of the component method, as explained later
in this paper.

Identication of the various components that constitute a joint (bolts, welds, stieners) gives a good picture of the complexity of its analysis, which requires
proper consideration of a multitude of phenomena,
ranging from material non-linearity (plasticity, strainhardening), non-linear contact and slip, geometrical
non-linearity (local instability) to residual stress conditions and complicated geometrical congurations. Although numerical approaches using non-linear nite
elements may deal with all these complexities, they require lengthy procedures and are very sensitive to the
modelling and analysis options. In practical terms, a
predictive approach must thus be based on simpler
models that eliminate much of the variability arising
from the analysis procedure itself. The so-called component method corresponds precisely to a simplied
mechanical model composed of extensional springs and
rigid links, whereby the joint is simulated by an appropriate choice of rigid and exible components. These
components represent a specic part of a joint that,
dependent on the type of loading, make an identied
contribution to one or more of its structural properties [17], as illustrated in Fig. 2a. Typical examples of

Fig. 2. Component method applied to a typical beam-to-column joint: (a) component model; (b) equivalent rotational spring.

518

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

components for bolted steel joints are (i) column web

panel in shear, (ii) end-plate in bending, (iii) column
ange in bending, (iv) beam web in tension, (v) column
web in compression, (vi) column web in tension, (vii)
beam ange and web in compression, (viii) bolts in tension and (ix) welds. In general, each of these components
is characterised by a non-linear forcedeformation
curve, although simpler idealisations are possible.
Several alternative spring and rigid link models have
been proposed [6], which share the same basic components. In the following, the simplied component model
of the revised Annex J of EC3 (1998) will be selected,
that, for simplicity, combines the bending behaviour of
the joint with the shear behaviour of the column panel to
yield an equivalent rotational spring, as shown in Fig.
2b.
Application of the component method to steel joints
requires the following steps:
(i) selection of the relevant (active) components from
a global list of components (13 dierent components
currently codied, for example, in Annex J of EC3);
(ii) evaluation of the forcedeformation response of
each component;
(iii) assembly of the active components for the evaluation of the momentrotation response of the joint,
using a representative mechanical model (Fig. 2a).
Its application may correspond to dierent levels of
renement, simplied characterisation of the components being possible whenever only the resistance or the
initial stiness of the joint is required.

2.2. Component characterisation

Describing the mechanical behaviour of the various
components of a joint allows the analysis of a large
number of dierent joint congurations with a relatively
small number of repeating components. A key aspect to
the component method thus relates to the characterisation of the forcedeformation curves for each individual
extensional spring. For the evaluation of the initial
stiness of a joint, only the linear stiness of each
component is required, whereas the evaluation of ductility requires the knowledge of the non-linear force
deformation response of each component.
Concentrating on the components relevant for steel
beam-to-column joints, a brief review of their behaviour
is presented below. With reference to Fig. 3a, it is noted
that analytical expressions are only presented for
strength and initial stiness because of lack of data for
the post-limit response, here presented only in a qualitative way, according to research results from various
authors. Of particular relevance to a ductility evaluation
is the deformation capacity of each component. Here, in

Fig. 3. Components with high ductility: (a) actual behaviour;

and (b) bi-linear approximation.

direct analogy with the classication of cross-sections,

three classes are proposed [10], described below.
2.2.1. Components with high ductility
According to Kuhlmann et al. [10], these components
present a forcedeformation curve that changes from an
initial linear elastic mode into a second carrying mode
which allows increasing deformation with increasing
force. The deformation capacity of the component is
nearly unlimited, not imposing any bounds on the
overall rotation ability of the joint, and is typically illustrated in Fig. 3a or, as a bi-linear approximation, in
Fig. 3b, where K e , K pl , F y and Dy denote, respectively,
the initial elastic stiness, the post-limit stiness, the
strength and the yield displacement of the component. It
is noted that Df , the limit displacement of the compo-

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

nent, is very high, so that, in practical terms, Df =Dy may

be taken as innity. Some components falling into this
classication are described below:
2.2.1.1. Column web panel in shear. This component has
been studied by Jaspart [8], typical experimental results
being reproduced in Fig. 4, that clearly show the stable
post-limit response.
The resistance of the panel zone in shear is given by
Vwp;Rd

fy;wc Awc
p
3

where fy;wc is the yield stress of the column web and Awc
is the shear area of the column. In case of welded sections, the shear area of the column coincides with the
area of the web, whereas in the case of rolled sections it
is given by
Awc Ac  2bc tfc twc 2rc tfc

where Ac is the total area of the column, bc , tfc and twc

are, respectively, the ange width, the ange thickness
and the web thickness of the column and rc is the rootradius of the web-ange junction. Eq. (1) neglects the
column axial load; otherwise, using the Von Mises yield
criterion it would be possible to evaluate a reduced value
of resistance that takes the column axial load into consideration. Jaspart [8] suggested a reduction coecient
of 0.9 that approximately takes care of this problem, an
approach currently adopted in Annex J of EC3, yielding
Vwc;Rd

0:9fy;wc Awc
p
3

519

According to Janss and Jaspart [7], Jaspart [8] and

Shi et al. [16], the contribution of the shear deformation
of the column web panel to the overall initial rotation of
the joint is given by
Us

Q
GAwc

where Q denotes
the shear force on the column web,
P
taken as 2 Fi (Fi denoting the force in each bolt row
and i the bolt row), and Awc already dened above. The
corresponding axial stiness becomes
Kwp

GAwc 0:38EAwc

z
z

where z denotes the lever arm between the compressive

and the tensile areas. From Eq. (4) it can be observed
that the stiness of this component depends on the applied shear force on the column web. Given that, in
general, internal forces transmitted by the lower and
upper column and (for internal nodes with unbalanced
moments) left beam may also be present, the applied
shear force must also be modied by a factor b to deal
with this eect, Eq. (5) becoming
Kwp

0:38EAwc
bz

For a stiened web panel the shear deformation may

be neglected (Ks;wp 1). Finally, it should be noted that
for slender webs, instability becomes the governing
factor, currently not covered in code specications.
2.2.1.2. End-plate in bending. The deformation of this
component is usually evaluated using a simple substitute
model, the T-stub [18,19], assumed to represent the behaviour of the tension zone of the joint and illustrated in
Fig. 5a. In terms of resistance, the T-stub exhibits three

Fig. 4. Experimental results taken from Jaspart [8].

Fig. 5. Equivalent T-stub assembly.

520

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

alternative failure modes, typically shown in Fig. 5b,

described next:
(i) Type 1end-plate yielding without bolt failure.
(ii) Type 2simultaneous yielding of end-plate with
bolt failure.
(iii) Type 3bolt failure without end-plate yielding.
Eq. (7) describes the corresponding axial strength,
8
4Mpl;Rd
>
>
Type 1
>
< m
P
2M

n
B
Ft;Rd
7
pl;Rd
t;Rd
Type 2
>
>
>
mn
:P
Bt;Rd
Type 3

2.2.1.3. Column ange in bending. Except for the restraint provided by additional stiening of the column,
this component behaves similarly to the end-plate in
bending, the T-Stub approach being equally valid. The
same degree of ductility and post-limit stiness is thus to
be expected, the relevant equations for strength and
stiness being reproduced below.
8
4Mpl;Rd
>
>
Type 1
>
< m
P
2M

n
B
Ft;Rd
10
pl;Rd
t;Rd
Type 2
>
>
>
mn
:P
Bt;Rd
Type 3
Kb;f

where m denotes the distance between the bolt centreline and the face of the weld connecting the beam web to
the end-plate, n is the eective distance to the free edge,
Bt;Rd corresponds to the resistance of the bolts in tension
and Mpl;Rd is the exural resistance of the end-plate,
given by
Mpl;Rd

leff tp2 fy;p

4

where leff is the eective width of the end-plate in

bending, and tp and fy;p are the thickness and yield stress
of the end-plate, respectively.
Analytical expressions for the initial stiness of the
T-Stub (end-plate in bending) can be derived from
classical beam theory [16,18], once an eective width has
been properly evaluated, giving
Kb;p

0:85Eleff tp3
m3

typical forcedeformation results obtained from experimental work being reproduced in Fig. 6 [5], showing a
stable (positive) post-limit stiness.

Fig. 6. Force versus deformation for T-Stub assembly taken

from Gebbeken et al. [5].

0:85Eleff tfc3
m3

11

the various quantities having the same meaning as

for the end-plate in bending, just replacing the end-plate
for the column ange.
2.2.1.4. Beam web in tension. For bolted end-plate joints,
the tension resistance of the beam web is given by
Ft;wb;Rd beff;t;wb twb fy;wb

12

where the eective width beff;t;wb should be taken as equal

to the eective length of the equivalent T-Stub representing the end-plate in bending and twb and fy;wb denote,
respectively, the thickness of the beam web and the
corresponding yield stress. The initial stiness for this
component may be taken as innity (Kt;wb 1).
2.2.2. Components with limited ductility
These components are characterised by a forcedeformation curve exhibiting a limit point and a subsequent
softening response, as shown in Fig. 7a or, as a bi-linear
approximation, in Fig. 7b. In this ductility class, it is
required to dene the collapse displacement of the
component, Df .
2.2.2.1. Column web in compression. This component has
been studied by Kuhlmann [9], who concluded that it
exhibited limited ductile behaviour with a softening
branch after reaching its maximum load carrying capacity, as reproduced in Fig. 8.
The resistance of this component may be subdivided
into two dierent criteria, crushing and buckling resistance. The crushing resistance must take into account
the interaction between local stresses that arise from the
shear stresses in the panel zone, the vertical normal
stresses due to axial load and bending moment in the
column and the horizontal normal stresses transmitted
by the beam anges. Using the Von Mises yield criterion
[4], the crushing resistance is given by
Fc;wc;Rd beff;c;wc twc fy;wc xkc;wc

13

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

521

where beff;c;wc is the eective width of the column web in

compression, given by, for bolted end-plate joints,
p
14
beff;c;wc tfb 2 2a 5tfc s sp
a denoting the eective thickness of the weld, s r for
rolled column sections and sp denoting the length obtained by dispersion at 45 through the end-plate; kc;wc
accounts for the inuence of vertical normal stress, rv ,
kc;wc 1:25  0:5

rv
6 1 rv > 0:5fy;wc
fy;wc

15

and x accounts for the shear interaction, given by [8]

8
0:0 6 b 6 0:5
<1
x x1 21  b1  x1 0:5 6 b 6 1:0
16
:
x1 b  1x2  x1 1:0 6 b 6 2:0
with
1
x1 q
1 1:3beff;c;wc twc =Avc 2

17a

1
x2 q
1 5:2beff;c;wc twc =Avc 2

17b

The buckling resistance is taken approximately using

the Winter formula as
Fc;wc;Rd 6 qbeff;c;wc twc fy;wc xkc;wc

Fig. 7. Components with limited ductility: (a) actual behaviour;

and (b) bi-linear approximation.

18

where q denotes the reduction factor for plate buckling,

given by
(
1
k < 0:673
19
q k0:22 
k > 0:673
2
k
and k denotes the normalised plate slenderness,
s
s
beff;c;wc twc fy;wc
beff;c;wc d fy;wc

k
0:932
2
Fcr
Etwc

20

The initial deformation of this component, Uc , may

be determined from [16]
Uc

N d
EAc hc

21

where
P N is the resultant compressive force, taken as
2 Fi (Fi denote the force in each bolt row and i the bolt
row), Ac is the eective web area in compression zone,
Ac twc beff;c , d the depth between column llets, and hc
the beam depth minus beam ange thickness, so that the
initial (axial) stiness becomes
Fig. 8. Forcedeformation response of column web in compression [9].

Kc;wc EAc

1
0:7beff;c;wc twc
E
d
d

22

522

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

where it is noted that for the stiness calculation a reduction of the eective width used for the strength calculation is adopted (0.7beff;c;wc ).
2.2.2.2. Column web in tension. Excluding instability
phenomena, the resistance of this component is similar
to the column web in compression. Consequently,
Ft;wc;Rd beff;t;wc twc fy;wc x kt;wc

23

where the various quantities take the same meaning as

before by replacing c for t. It is noted that Annex J of
EC3 disregards the inuence of vertical stresses arising
from the column.
In analogy with the previous case, the initial deformation of this component, Uw , may be determined from
[16]
Uw

T d
EAt ht

24

P
where T is the resultant tensile force, taken as 2 Fi (Fi
denoting the force in each bolt row and i the bolt row),
At is the eective web area in the tensile zone, At
twc beff;t;wc , d the depth between column llets, and ht the
distance from the tensile force to the center of compression, so that the axial stiness becomes
Kt;wc EAt

1
0:7beff;t;wc wc
E
d
d

25

2.2.2.3. Beam ange and beam web in compression. The

beam ange and web in compression adjacent to
the beam-connection system provides a limitation to
the resistance of the joint, so that it is required to assess
its maximum resistance, given by
Fc;fb;Rd

Mc;Rd
z

26

Fig. 9. Components with brittle failure: (a) actual behaviour;

and (b) linear approximation.

while its initial stiness is taken as innity.

2.2.3. Components with brittle failure
These components behave linearly until collapse,
with very little deformation before failure, as shown in
Fig. 9a or, as a linear approximation, in Fig. 9b, so that
Df Dy .
2.2.3.1. Bolts in tension. Bolts exhibit a linear force
deformation response up to failure, as shown in Fig. 10,
taken from a tensile test on a single bolt. The resistance
and initial stiness of each bolt are given by
Ft;Rd 0:9fub As
Kt;b

1:6EAs
Lb

27
28

where As is the tensile area of the bolt, fub the ultimate

tensile strength of the bolts and Lb is the sum of the

Fig. 10. Experimental results for bolt in tension.

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

thickness of the connected plates, the thickness of the

washers and the half thickness of the nut and the bolt
head.
2.2.3.2. Welds. Welds are virtually undeformable (Kw
1), a rigid-plastic model being adequate, resistance
being given by
p
fu = 3
29
Fw;Rd a
bw
where a is the eective thickness of the weld, fu the ultimate tensile strength of the weld and bw is a correlation
factor.

3. Joint ductility
The assessment of the ductility of a steel joint requires
a non-linear procedure, which takes into account the
non-linear forcedeformation response of each component. Here a bi-linear forcedeformation response with a
cut-o is assumed which highlights, for each component, the transition between initial elastic stiness and
residual stiness while maintaining sucient accuracy.
Additionally, direct comparison with ideal linear elastic
components is straightforward using, for example, the
values for component stiness that were presented above.

4. Post-limit stiness of bolted end-plate beam-to-column

joints
4.1. Introduction
In order to evaluate realistic values of the post-limit
stiness of the various relevant components, a set of four
extended end-plate beam-to-column joints tested by
Humer at the University of Innsbruck (1987) were selected from the database of steel joints SERICON II [1].

523

Table 1
Mechanical properties of Humer tests
Elements

Yield strength
(MPa)

Humer 109.005
Column web
Column ange
Beam web
Beam ange
End-plate
Backing plate
Boltsa

306.6
275.9
315.6
284.6
323.0
333.0
900.0

Humer 109.006
Column web
Column ange
Beam web
Beam ange
End-plate
Backing plate
Boltsa

309.4
247.6
294.2
288.0
325.0
336.0
900.0

Humer 109.003
Column web
Column ange
Beam web
Beam ange
End-plate
Backing plate
Boltsa

341.1
300.4
343.8
322.5
273.0
368.0
900.0

Humer 109.004
Column web
Column ange
Beam web
Beam ange
End-plate
Backing plate
Boltsa

359.0
305.9
315.6
284.6
323.0
298.0
900.0

Failure strength
(MPa)
445.0
400.5
398.0
413.0
360.0
1000.0
449.0
398.5
427.0
418.0
360.0
1000.0
499.0
468.0
458.0
413.0
360.0
1000.0
521.0
444.0
458.0
413.0
360.0
1000.0

E 210 GPa.
a
Nominal values.

For all specimens, the (measured) material properties

and geometries are reproduced in Tables 1 and 2, the

Table 2
Geometric properties of Humer tests
Test
109.005

109.006

109.003

109.004

Beam

IPE450R: h 454;
b 192; tf 14;
tw 10:4; r 21

IPE600R: h 597;
b 220; tf 18:6;
tw 12:1; r 24

IPE300R: h 300;
b 151; tf 11:2;
tw 7; r 15

IPE450R: h 454;
b 192; tf 14:6;
tw 10:4; r 21

Column

HEB240R: h 242;
b 240; tf 16:4;
tw 10:4; r 21
553 239 41
2 185 100 12
3 bolt rows M24

HEB240R: h 240;
b 240; tf 17;
tw 10:4; r 21
693 243 40
2 200 95 12
3 bolt rows M24

HEB180R: h 179;
b 180; tf 14:1;
tw 9:2; r 15
383 181 30
2 150 70 10:8
3 bolt rows M20

HEB180R: h 180;
b 180; tf 14;
tw 9:4; r 15
553 239 41
2 185 100 10:4
3 bolt rows M24

End-plate
Backing-plate
Bolts
Units: mm.

524

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

Fig. 11. Joint geometry for Humer 109.005.

Fig. 13. Experimental results for test Humer 109.005.

Fig. 12. Localisation of displacement transducers.

layout of the joint, instrumentation and corresponding

experimental results for test 109.005 being illustrated in
Figs. 1113.
For each specimen, a prediction of the moment
rotation response was attempted using the component

model of Fig. 2a. This component model was analysed

either by applying the analytical methodology presented
in [14] or by performing a non-linear nite element analysis using the bi-linear characteristics of the components.
The adopted procedure for establishing the post-limit
stiness of the various components involves the following steps:
(a) for each specimen, assumption of trial values of
the post-limit stiness, obtained as a best t to the experimental momenttotal rotation curves;

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

525

(b) for each specimen, and where available, best t

calibration of component sub-models with experimental curves for moment versus panel rotation
and moment versus connection rotation;
(c) statistical evaluation (mean and standard deviation) of the normalised post-limit stiness values obtained above (ratio of initial stiness versus post-limit
stiness) for steps (a) and (b);
(d) for each specimen, evaluation of momentrotation curves for the average values established above.

5. Numerical models
The numerical model adopted in the analysis for the
chosen joint congurations are illustrated in Fig. 14. The
rigid links are modelled using beam elements with elastic
material properties and very high cross-sectional properties, while the springs are modelled as non-linear joint
elements, reproducing the bi-linear characteristics earlier
described. An incremental non-linear analysis for an
applied bending moment is performed using the nonlinear nite element code [12].
According to the procedure dened above, distinct
numerical models were dened for step (a) (Fig. 15a:
moment versus total joint rotation) and step (b) (Fig.
15b (1): moment versus panel rotation and 15b (2):
moment versus connection rotation).

Fig. 15. Component method model: (a) joint model; and (b) (1)
shear panel model, (2) connection model.

6. Results and discussion

Fig. 14. Finite element model: Humer 109.005.

Starting, for exemplication, with test 109.005, Table

3 reproduces the strength and initial stiness values for
all relevant components. Application of step (a) of the
procedure described above leads to the results of Fig. 16,
that compares the experimental results with the numerical/analytical results, showing excellent agreement between both curves. For the calculated momentrotation
curve, the yielding rotations of the critical components
are also identied. Similarly, application of step (b)
yields the results of Fig. 17a and b, that compare the
experimental and numerical curves for moment versus
panel rotation and moment versus connection rotation,
respectively.
Repeating steps (a) and (b) for the remaining joint
congurations and dening the normalised post-limit

526

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

Table 3
Component characterisation for Humer 109.005
Component

Designation

F y (kN)

k (mm)

K e (kN/m)

Dy (mm)

Column web in shear

Column web in compression
Column web in tension

1
2
3.1
3.2
4.1
4.2
5.1
5.2
7
8.1
8.2
10.1
10.2

543.83
602.31
386.24
386.24
435.93
435.93
635.40
635.40
877.10
941.21
941.21
635.40
635.40

2.91
11.71
7.12
7.12
17.85
17.85
114.60
116.87
1
1
1
6.25
6.25

611100
2459100
1495200
1495200
3748500
3748500
24066000
24542700
1
1
1
1312500
1312500

0.88992
0.24493
0.25832
0.25832
0.11629
0.11629
0.02640
0.02589

Column ange in bending

End-plate in bending
Beam web in tension
Beam ange in compression
Bolts in tension

0.48411
0.48411

Fig. 16. Momentrotation curve Humer 109.005 joint.

stiness as the ratio, expressed as a percentage, between

the post-limit stiness and the corresponding initial
stiness,

Ki

Kipl
100
Kie

30

leads to the results of Table 4, that illustrates the calibrated values for the critical components, together with
the statistical evaluation (mean and standard deviation)
of the normalised post-limit stiness for each component
(step (c)).
Examination of the normalised post-limit stiness
values of Table 4 led to the choice of average values for
the various components shown in Table 5. Assuming
these mean values for all specimens, and reanalysing all
cases using these properties (step (d)) yields the results of
Figs. 16,1820, where the experimental results are plotted
superimposed with the numerical results earlier obtained
by individual calibration of the post-limit stiness values

Fig. 17. Momentrotation curve Humer 109.005: (a) shear

panel rotation; and (b) connection rotation.

(exact solution, step (a)) and the corresponding results

obtained using the average values of post-limit stiness.
In order to assess the error of this approach, an adimensional error measure is proposed, given by Eq. (31),
s
n
X
p
e
ei ef
i1

31

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

527

Table 4
Calibrated values of post-limit stiness
Elastic Stiness
(kN/m)

Component 1

Humer
Humer
Humer
Humer

109.005
109.006
109.003
109.004

Momenttotal joint rotation

Momentshear panel/connection rotation

Post-limit stiness (kN/m)

Post-limit stiness (kN/m)

611100
468300
573300
388500

48000
20000
27138
16535

7.86%
4.27%
4.73%
4.26%

Mean
Standard deviation
Component 2

Humer
Humer
Humer
Humer

109.005
109.006
109.003
109.004

2459100
2553600
2362500
2545200

95000
22000
92269
23622

Humer
Humer
Humer
Humer

1495200
1635900

16988
3043

70000
34048
90865
29944

3.05%
1.33%
3.85%
1.17%
2.35%
0.013

1.14%
0.19%

41073
42328

2.75%
2.58%

0.66%
0.007

109.005
109.006
109.003
109.004

3748500
2499000
2706900
1409100

88184
14366
141912
4464

Humer 109.005
Mean

1495200

1048

Table 5
Adopted normalised post-limit stiness values for the various
components
Component

K i (%)

1
2
3.1
3.2
4.1
10.1
10.2

4.59
2.35
1.67
0.10
1.31
7.19
0.45

2.67%
0.001

2.35%
0.58%
5.24%
0.32%

Mean (without
Humer 109.003)
Standard deviation
Component 3.2

5.40%
4.91%
4.17%
3.86%

2.39%
0.017

Humer 109.005
Humer 109.006

Ki

4.59%
0.007

3.86%
0.86%
3.91%
0.93%

Mean
Standard deviation
Component 4.1

33000
23000
24000
15000

5.28%
0.016

Mean
Standard deviation
Component 3.1

Ki

52275
50023
331128
7603

1.40%
2.00%
12.23%
0.54%

1.08%

1.31%

0.023

0.055

0.07%
0.07%

1442

0.10%
0.1%

Mij denoting the moment at yield of component i,

superscript j ind; av denoting individual calibration
(step (a)) or use of average values (step (d)). Similarly, /i
denotes the corresponding joint rotation at yield of
component i. The second term in Eq. (31) estimates
(where applicable) the error at failure of the joint, given
by


ef

Mfind  Mfav
Mfind

/if  /av
f
/ind
f

!2
33

where

ei

Miind  Miav
Miind

/ind
 /av
i
i
/ind
i

!2
32

Mf and /f having the same meaning as before, subscript

f denoting failure of the joint. Table 6 illustrates the
error for each test.

528

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

Table 6
Error evaluation for teach test result
Humer

Error (%)

109.005
109.006
109.003
109.004

7.871
3.863
1.555
1.857

7. Component ductility index

7.1. Denition

Fig. 18. Humer 109.006: momentrotation curve, experimental

results and analytical response.

The evaluation of the ductility of steel joints in the

context of the component method requires, as mentioned above, the characterisation of the ductility of
each component, i.e., the identication of the failure
displacement, Df , of each component. Here, assuming
the bi-linear idealisation of component behaviour of Fig.
7b, a ductility index ui is proposed for each component
i, dened as,
ui

Dfi
Dyi

34

The component ductility index ui allow a direct

classication of each component in terms of ductility,
using, for example, the three ductility classes proposed
by Kuhlmann et al. [10]:
Class 1components with high ductility ui P a:
Class 2components with limited ductility b 6 ui < a:
Class 3components with brittle failure ui < b:
Fig. 19. Humer 109.003: momentrotation curve, experimental
results and analytical response.

a and b representing ductility limits for the various

component classes, here suggested as a 20 and b 3.
In design terms, and in-line with the usual assumptions
in plastic design, it seems reasonable to assume, for
Class 1 components, a ductility index ui 1. On the
other end, for Class 3 components, because of brittle
behaviour, a safe estimate can be obtained with a ductility index of ui 1 (elastic response). For Class 2
components, lower bounds for the ductility indexes must
be established for each component type, as a result of
experimental and analytical research to be carried out.
As a crude indication, from the experimental results
obtained by Kuhlmann [9] and referring to Fig. 8, a
ductility index in the range of 45 seems reasonable for
the component web in compression, if a negative plastic
stiness is used.
7.2. Application to end-plate beam-to-column joints

Fig. 20. Humer 109.004: momentrotation curve, experimental

results and analytical response.

Evaluation of the ductility indexes for test 109.005

yields the results of Table 7. Examination of Table 7

Rel. displ.

2.3687
1.0000
0.8811
0.6783
0.7807
0.6010
0.5393
0.4093
0.5356
0.4123
2.3889
3.7941
1.0969
0.2583
0.1987
0.1030
0.0792
0.0160
0.0121
0.2943
0.2263
0.0189
2.1079
0.2449
0.2276
0.1752
0.0908
0.0699
0.0141
0.0107
0.2593
0.1996
0.0086

Abs. displ. Di

0.8899
0.2211
0.2056
0.1581
0.0820
0.0631
0.0128
0.0096
0.2342
0.1801
0.0036
Abs. rot. (radian)
INF
5
INF
INF
INF
INF
INF
INF
1
1

1
2
3.1
3.2
4.1
4.2
5.1
5.2
10.1
10.2
Joint
Rotation

Component yield sequence (j)

ui Dfi =Dyi
Component

Table 7
Ductility indexes for extended end-plate joint 109.005

5.6194
2.0192
3.1832
0.2241
0.1163
0.0894
0.0181
0.0137
0.3321
0.2553
0.0326

8.0380
1.0000
3.2412
0.9028
7.0365
0.7960
0.2583
0.6121
0.1338
0.7051
0.1030
0.5426
0.0208
0.4896
0.0157
0.3672
0.3820
0.4838
0.2943
0.3720
0.0565
1.0000
Rel. rot. h/h1

Di =Dyi

4.2635
4.4790
1.0000
0.7693
0.8856
0.6810
0.6120
0.4628
0.6079
0.4675
5.2500

6.3146
8.2450
12.324
0.8676
1.0000
0.7687
0.6923
0.5240
0.6860
0.5274
9.0556

9.0325
13.235
27.242
1.0000
1.1505
0.8856
0.7956
0.6005
0.7891
0.6079
15.689

Failure

9.0325
13.235
27.242
1.0000
1.1505
0.8856
0.7956
0.6005
0.7891
0.6079
15.689

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

529

clearly shows the yield sequence of the various components and the corresponding levels of ductility for the
analysed extended end-plate joint. As also observed in
Fig. 16, the rst component to yield is the column web in
shear, at a yield displacement of 0.8899 mm (Table 7)
and a total joint rotation of 0.0036 radian, the other
components remaining elastic. Next, in succession, the
following components reach yield: column web in
compression (2), column web in tension (3.1), column
ange in bending (4.1) and column web in tension (3.2).
Table 7 illustrates the relevant values of displacement
and the corresponding values for the remaining components. Finally, for this test, the maximum recorded
value of total rotation was 0.056 radian.
A joint ductility index can also be proposed, dened
as
uj

hf
h1

35

where hf denotes the rotation at failure and h1 the rotation when the rst component reaches its elastic limit.
For the four examples presented above, the joint ductility index varies between 5 and 43, based on the maximum experimentally recorded rotation for each test. It
is noted that except for test 109.003, no brittle components reached yield, casting some doubts over the likelihood of this particular result, since no sudden failure of
the joint was subsequently observed. Also of importance
is the maximum ductility index reached by the components with limited ductility, a maximum ratio of 46 being calculated for the column web in compression
without failure.

8. Conclusions
The evaluation of the ductility of a steel joint within
the scope of the component method requires proper
characterisation of each component. A good balance
between relative simplicity and rigorous results may be
achieved using bi-linear approximations of the force
deformation behaviour of each component, including
the post-limit stiness. Because many components are
still not adequately characterised, work remains to be
done in that area before ductility indexes can be established for each component that, dependant on its geometric and material properties, correspond to safe
estimates of deformation ability for each component.
This explains some less plausible results for the yield
sequence of the various components that may arise from
a certain conservative evaluation of the yield strength of
some components. A good example of such a situation is
the column web in compression, improved expressions

530

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

Table 8
Proposed code coecients for the evaluation of rotation capacity
Component

Application rules
Resistance

Rotation capacity

0:9f
py;wc
Avc
3cM0
xb
twc fy;wc
eff;c;wc
cM0

Initial stiness

Post-limit stiness K i

Limit displacement

[4.6%]

Df1 1

1 Column web panel in shear

Vwp;Rd

vc
K1 E 0:38A
bz

2 Column web in compression

Fc;wc;Rd

K2 E

0:7beff;c;wc twc
dc

[2.3%]

Df2 3  5Dy2

K3 E

0:7beff;t;wc twc
dc

[0.11.7%]

Df3 1

K4 E

3
0:85leff tfc
m3

[1.3%]

Df4 1

K5 E

0:85leff tp3
m3

Df5 1

K7 1

Df7 3  5Dy7

K8 1
s
K10 E 1:6A
Lb

a
Components with
brittle failure
Components with
brittle failure
Components with
brittle failure

Df8 1
Df10 Dy10

but Fc;wc;Rd 6
3 Column web in tension
4 Column ange in bending
5 End-plate in bending
7 Beam or column ange and
web in compression
8 Beam web in tension
10 Bolts in tension
11 Bolts in shear
19 Welds
a

Ft;wc;Rd

xqbeff;c;wc twc fy;wc

cM1

xbeff;t;wc twc fy;wc

cM0

Equivalent T-stub model

[Annex J-EC3, J.3.2]
Equivalent T-stub model
[Annex J-EC3, J.3.2]
M
Fc;fb;Rd c;Rd
z
b

Ft;wb;Rd eff;t;wbcM0wb
ub As
Ft;Rd 0:9fcMb

fy;wb

Varying according to bolt

grade
p
Fw;Rd a bfuw=cMw3

d fub
K11 E 16nEdbM16

K19 1

Df11 Dy11
Df19 Dy19

Values to be established.

for its resistance being recently proposed by Kuhlmann

and Kuenhemund [11].
The current draft version of Part 1.8 of EC3 [3] already tries to extend the vague ductility provisions that
were present in Annex J of EC3 by presenting a table
with an unlled column for rotation capacity, component by component. Table 8 presents an improved version of this table which includes two columns for
rotation capacity: post-limit stiness and limit displacement. This subdivision is required since no ductility
limits may be evaluated without the prior knowledge of
a post-limit stiness [13]. Based on the statistical analysis performed in this paper for a limited number of
test results (single-sided, extended end-plate beam-tocolumn joints with backing plates between an IPE beam
and a HEB column), some trial values are proposed (in
brackets) as a rst approximation.
Next, a ductility model is required which is able to
predict the yield sequence of the various components
[14] and a safe (lower bound) joint ductility index, here
chosen as a relative value of total rotation with respect
to the initial stiness of the joint.
Finally, it is worth pointing out that ductility evaluation should be performed on actual values of component behaviour (particularly when talking in terms of
strength), because of the unexpected results of over-

strength eects that may produce unsafe results [15].

This may even lead to the requirement of guaranteed
upper bounds on material properties, in particular for
the yield stress of steel.

Acknowledgements
Financial support from Ministerio da Ci^encia e
TecnologiaPRAXIS XXI research project PRAXIS/P/
ECM/13153/1998 is acknowledged.

References
[1] Cruz PJS, Silva LS, Rodrigues DS, Sim~
oes R. Database for
the semi-rigid behaviour of beam-to-column connections in
seismic regions. J Construct Steel Res 1998;46(120):13.
[2] CEN. Eurocode 3, ENV-1993-1-1, Revised Annex J,
Design of Steel Structures, CEN, European Committee
for Standardization, Document CEN/TC 250/SC 3-N 419
E, Brussels, 1998.
[3] CEN. Eurocode 3, prEN-1993-1-8: 20xx, Part 1.8: Design of
Joints, Eurocode 3: Design of Steel Structures, Draft 2 Rev.,
6 December 2000, CEN, European Committee for Standardization, Brussels, 2000.

L. Sim~oes da Silva et al. / Computers and Structures 80 (2002) 515531

[4] Faella C, Piluso V, Rizzano G. Structural steel semirigid
connections: theory, design and software. Boca Raton, FL:
CRC Press; 2000.
[5] Gebbeken N, Wanzek T, Petersen C. Semi-rigid connections T-stub model. M
unchen, Germany: Institut f
ur
Mechanik und Statik, Universitat des Bundeswehr
M
unchen; 1997.
[6] Huber G, Tschemmernegg F. Modelling of steel connections. J Construct Steel Res 1998;45(2):199216.
[7] Janss J, Jaspart JP. Strength and behaviour of in plane
weak axis joints and of 3-D joints. In: Bjorhovde R, Colson
A, Zandonini R, editors. Connections in steel structures.
Proceedings of the International Workshop on Joints.
New York: Elsevier Applied Science; 1987.
[8] Jaspart JP. Etude de la semi-rigidite des noeuds poutrecolonne et son inuence sur la resistance des ossatures en
acier. PhD thesis, Department MSM, University of Liege,
Belgium, 1991 [in French].
[9] Kuhlmann U. Inuence of axial forces on the component:
web under compression. Proceeding of COST-C1 Working
Group Meeting, C1/WG2/99-01. Thessaloniki, May 1999.
[10] Kuhlmann U, Davison JB, Kattner M. Structural systems
and rotation capacity. Proceeding of COST Conference on
Control of the Semi-rigid Behaviour of Civil Engineering
Structural Connections, Liege, Belgium, 1998. p. 16776.
[11] Kuhlmann U, Kuhnemund F. Proposal of a new design
resistance of the joint component column web in compression, Internal Report. Stuttgart, Germany: University of
Stuttgart; January 2001.

531

[12] LUSAS. Lusas Finite Element System, LusasUser Manual, Version 13.3. FEAFinite Element Analysis Ltd.,
Kingston-upon-Thames, England, 2000.
[13] Silva LS, Coelho AG, Neto EL. Equivalent post-buckling
models for the exural behaviour of steel connections.
Comput Struct 2000;77:61524.
[14] Silva LS, Coelho AG. A ductility model for steel connections. J Construct Steel Res 2001;57:4570.
[15] Silva LS, Gervasio H, Rebelo C, Coelho AG. Assessment
of overstrength eects in steel and composite connections
using Monte Carlo methods. In: Proceedings of IABSE
International Conference on Safety, Risk and Reliability
Trends in Engineering, Malta, 2123 March 2001. p.
22930.
[16] Shi YJ, Chan SL, Wong YL. Modelling for moment
rotations characteristics for end-plate joints. J Struct
Engng 1996;122(11):13006.
[17] Weynand K, Jaspart JP, Steenhuis M. The stiness model
of revised Annex J of Eurocode 3. In: Bjorhovde R, Colson
A, Zandonini R, editors. Connections in steel structures
III. Proceedings of the 3rd International Workshop on
Connections in Steel Structures. Trento, Italy, 1995. p.
44152.
[18] Yee YL, Melchers RE. Momentrotation curves for
bolted connections. J Struct Engng ASCE 1986;112(3):
61535.
[19] Zoetemeijer P. A design method for the tension side of
statically-loaded bolted beam-to-column joints. Heron
1974;20(1):159.